Search arXiv⌕ Search

arXiv subjects

Mingyang Wang

Publications and source records attributed to Mingyang Wang.

At least 37 records · Page 2Linked to original sources
arXiv2026

SAD: A Large-Scale Strategic Argumentative Dialogue Dataset

Argumentation generation has attracted substantial research interest due to its central role in human reasoning and decision-making. However, most existing argumentative corpora focus on non-interactive, single-turn settings, either generating arguments from a given topic or refuting an existing argument. In practice, however, argumentation is often realized as multi-turn dialogue, where speakers defend their stances and employ diverse argumentative strategies to strengthen persuasiveness. To support deeper modeling of argumentation dialogue, we present the first large-scale \textbf{S}trategic \textbf{A}rgumentative \textbf{D}ialogue dataset, SAD, consisting of 392,822 examples. Grounded in argumentation theories, we annotate each utterance with five strategy types, allowing multiple strategies per utterance. Unlike prior datasets, SAD requires models to generate contextually appropriate arguments conditioned on the dialogue history, a specified stance on the topic, and targeted argumentation strategies. We further benchmark a range of pretrained generative models on SAD and present in-depth analysis of strategy usage patterns in argumentation.

cs.CL↗
arXiv2026

PlaM: Training-Free Plateau-Guided Model Merging for Better Visual Grounding in MLLMs

Multimodal Large Language Models (MLLMs) rely on strong linguistic reasoning inherited from their base language models. However, multimodal instruction fine-tuning paradoxically degrades this text's reasoning capability, undermining multimodal performance. To address this issue, we propose a training-free framework to mitigate this degradation. Through layer-wise vision token masking, we reveal a common three-stage pattern in multimodal large language models: early-modal separation, mid-modal alignment, and late-modal degradation. By analyzing the behavior of MLLMs at different stages, we propose a plateau-guided model merging method that selectively injects base language model parameters into MLLMs. Experimental results based on five MLLMs on nine benchmarks demonstrate the effectiveness of our method. Attention-based analysis further reveals that merging shifts attention from diffuse, scattered patterns to focused localization on task-relevant visual regions. Our repository is on https://github.com/wzj1718/PlaM.

cs.CL↗
arXiv2026

A PDE approach for the invariant measure of stochastic oscillators with hysteresis

This paper presents a PDE approach as an alternative to Monte Carlo simulations for computing the invariant measure of a white-noise-driven bilinear oscillator with hysteresis. This model is widely used in engineering to represent highly nonlinear dynamics, such as the Bauschinger effect. The study extends the stochastic elasto-plastic framework of Bensoussan et al. [SIAM J. Numer. Anal. 47 (2009), pp. 3374--3396] from the two-dimensional elasto-perfectly-plastic oscillator to the three-dimensional bilinear elasto-plastic oscillator. By constructing an appropriate Lyapunov function, the existence of an invariant measure is established. This extension thus enables the modelling of richer hysteretic behavior and broadens the scope of PDE alternatives to Monte Carlo methods. Two applications demonstrate the method's efficiency: calculating the oscillator's threshold crossing frequency (providing an alternative to Rice's formula) and probability of serviceability.

math-ph↗
arXiv2025

HyperVL: An Efficient and Dynamic Multimodal Large Language Model for Edge Devices

Current multimodal large lanauge models possess strong perceptual and reasoning capabilities, however high computational and memory requirements make them difficult to deploy directly on on-device environments. While small-parameter models are progressively endowed with strong general capabilities, standard Vision Transformer (ViT) encoders remain a critical bottleneck, suffering from excessive latency and memory consumption when processing high-resolution inputs.To address these challenges, we introduce HyperVL, an efficient multimodal large language model tailored for on-device inference. HyperVL adopts an image-tiling strategy to cap peak memory usage and incorporates two novel techniques: (1) a Visual Resolution Compressor (VRC) that adaptively predicts optimal encoding resolutions to eliminate redundant computation, and (2) Dual Consistency Learning (DCL), which aligns multi-scale ViT encoders within a unified framework, enabling dynamic switching between visual branches under a shared LLM. Extensive experiments demonstrate that HyperVL achieves state-of-the-art performance among models of comparable size across multiple benchmarks. Furthermore, it significantly significantly reduces latency and power consumption on real mobile devices, demonstrating its practicality for on-device multimodal inference.

cs.CV↗
arXiv2025

On the Entity-Level Alignment in Crosslingual Consistency

Multilingual large language models (LLMs) are expected to recall factual knowledge consistently across languages. However, the factors that give rise to such crosslingual consistency -- and its frequent failure -- remain poorly understood. In this work, we hypothesize that these inconsistencies may arise from failures in entity alignment, the process of mapping subject and object entities into a shared conceptual space across languages. To test this, we assess alignment through entity-level (subject and object) translation tasks, and find that consistency is strongly correlated with alignment across all studied models, with misalignment of subjects or objects frequently resulting in inconsistencies. Building on this insight, we propose SubSub and SubInj, two effective methods that integrate English translations of subjects into prompts across languages, leading to substantial gains in both factual recall accuracy and consistency. Finally, our mechanistic analysis reveals that these interventions reinforce the entity representation alignment in the conceptual space through model's internal pivot-language processing, offering effective and practical strategies for improving multilingual factual prediction.

cs.CL↗
arXiv2025

BlackboxNLP-2025 MIB Shared Task: Exploring Ensemble Strategies for Circuit Localization Methods

The Circuit Localization track of the Mechanistic Interpretability Benchmark (MIB) evaluates methods for localizing circuits within large language models (LLMs), i.e., subnetworks responsible for specific task behaviors. In this work, we investigate whether ensembling two or more circuit localization methods can improve performance. We explore two variants: parallel and sequential ensembling. In parallel ensembling, we combine attribution scores assigned to each edge by different methods-e.g., by averaging or taking the minimum or maximum value. In the sequential ensemble, we use edge attribution scores obtained via EAP-IG as a warm start for a more expensive but more precise circuit identification method, namely edge pruning. We observe that both approaches yield notable gains on the benchmark metrics, leading to a more precise circuit identification approach. Finally, we find that taking a parallel ensemble over various methods, including the sequential ensemble, achieves the best results. We evaluate our approach in the BlackboxNLP 2025 MIB Shared Task, comparing ensemble scores to official baselines across multiple model-task combinations.

cs.CL↗
arXiv2025

On Relation-Specific Neurons in Large Language Models

In large language models (LLMs), certain \emph{neurons} can store distinct pieces of knowledge learned during pretraining. While factual knowledge typically appears as a combination of \emph{relations} and \emph{entities}, it remains unclear whether some neurons focus on a relation itself -- independent of any entity. We hypothesize such neurons \emph{detect} a relation in the input text and \emph{guide} generation involving such a relation. To investigate this, we study the LLama-2 family on a chosen set of relations, with a \textit{statistics}-based method. Our experiments demonstrate the existence of relation-specific neurons. We measure the effect of selectively deactivating candidate neurons specific to relation $r$ on the LLM's ability to handle (1) facts involving relation $r$ and (2) facts involving a different relation $r' \neq r$. With respect to their capacity for encoding relation information, we give evidence for the following three properties of relation-specific neurons. \textbf{(i) Neuron cumulativity.} Multiple neurons jointly contribute to processing facts involving relation $r$, with no single neuron fully encoding a fact in $r$ on its own. \textbf{(ii) Neuron versatility.} Neurons can be shared across multiple closely related as well as less related relations. In addition, some relation neurons transfer across languages. \textbf{(iii) Neuron interference.} Deactivating neurons specific to one relation can improve LLMs' factual recall performance for facts of other relations. We make our code and data publicly available at https://github.com/cisnlp/relation-specific-neurons.

cs.CL↗
arXiv2025

Tracing Multilingual Factual Knowledge Acquisition in Pretraining

Large Language Models (LLMs) are capable of recalling multilingual factual knowledge present in their pretraining data. However, most studies evaluate only the final model, leaving the development of factual recall and crosslingual consistency throughout pretraining largely unexplored. In this work, we trace how factual recall and crosslingual consistency evolve during pretraining, focusing on OLMo-7B as a case study. We find that both accuracy and consistency improve over time for most languages. We show that this improvement is primarily driven by the fact frequency in the pretraining corpus: more frequent facts are more likely to be recalled correctly, regardless of language. Yet, some low-frequency facts in non-English languages can still be correctly recalled. Our analysis reveals that these instances largely benefit from crosslingual transfer of their English counterparts -- an effect that emerges predominantly in the early stages of pretraining. We pinpoint two distinct pathways through which multilingual factual knowledge acquisition occurs: (1) frequency-driven learning, which is dominant and language-agnostic, and (2) crosslingual transfer, which is limited in scale and typically constrained to relation types involving named entities. We release our code and data to facilitate further research at https://github.com/cisnlp/multilingual-fact-tracing.

cs.CL↗
arXiv2025

Language Mixing in Reasoning Language Models: Patterns, Impact, and Internal Causes

Reasoning language models (RLMs) excel at complex tasks by leveraging a chain-of-thought process to generate structured intermediate steps. However, language mixing, i.e., reasoning steps containing tokens from languages other than the prompt, has been observed in their outputs and shown to affect performance, though its impact remains debated. We present the first systematic study of language mixing in RLMs, examining its patterns, impact, and internal causes across 15 languages, 7 task difficulty levels, and 18 subject areas, and show how all three factors influence language mixing. Moreover, we demonstrate that the choice of reasoning language significantly affects performance: forcing models to reason in Latin or Han scripts via constrained decoding notably improves accuracy. Finally, we show that the script composition of reasoning traces closely aligns with that of the model's internal representations, indicating that language mixing reflects latent processing preferences in RLMs. Our findings provide actionable insights for optimizing multilingual reasoning and open new directions for controlling reasoning languages to build more interpretable and adaptable RLMs.

cs.CL↗
arXiv2025

Dense Matter in Neutron Stars with eXTP

In this White Paper, we present the potential of the enhanced X-ray Timing and Polarimetry (eXTP) mission to constrain the equation of state of dense matter in neutron stars, exploring regimes not directly accessible to terrestrial experiments. By observing a diverse population of neutron stars - including isolated objects, X-ray bursters, and accreting systems - eXTP's unique combination of timing, spectroscopy, and polarimetry enables high-precision measurements of compactness, spin, surface temperature, polarimetric signals, and timing irregularity. These multifaceted observations, combined with advances in theoretical modeling, pave the way toward a comprehensive description of the properties and phases of dense matter from the crust to the core of neutron stars. Under development by an international Consortium led by the Institute of High Energy Physics of the Chinese Academy of Sciences, the eXTP mission is planned to be launched in early 2030.

astro-ph.HE↗
arXiv2025

Discovery of the anti-glitch in PSR J1835$-$1106

We report the detection of an anti-glitch with a fractional frequency change of $Δν/ν=-3.46(6)\times10^{-9}$ in the rotation-powered pulsar PSR J1835$-$1106 at MJD 55813, based on timing observations collected with the Nanshan 26-m and Parkes 64-m radio telescopes from January 2000 to July 2022. A comparison of the average pulse profiles within $\pm300$ d of the event reveals no significant morphological changes. We also estimate the angular velocity lag between the normal and superfluid components at the time of the glitch, showing that one of the superfluid glitch models is incompatible with PSR J1835$-$1106 due to its insufficient spin-down rate and angular velocity lag. The wind braking scenario offers a viable alternative, consistent with the observed spin-down behavior, glitch amplitude, and post-glitch recovery. High-cadence, high-sensitivity monitoring of similar events is essential to distinguish between internal (superfluid) and external (wind-related) glitch mechanisms.

astro-ph.HE↗
arXiv2025

Multimodal Recommendation via Self-Corrective Preference Alignmen

With the rapid growth of live streaming platforms, personalized recommendation systems have become pivotal in improving user experience and driving platform revenue. The dynamic and multimodal nature of live streaming content (e.g., visual, audio, textual data) requires joint modeling of user behavior and multimodal features to capture evolving author characteristics. However, traditional methods relying on single-modal features or treating multimodal ones as supplementary struggle to align users' dynamic preferences with authors' multimodal attributes, limiting accuracy and interpretability. To address this, we propose MSPA (Multimodal Self-Corrective Preference Alignment), a personalized author recommendation framework with two components: (1) a Multimodal Preference Composer that uses MLLMs to generate structured preference text and embeddings from users' tipping history; and (2) a Self-Corrective Preference Alignment Recommender that aligns these preferences with authors' multimodal features to improve accuracy and interpretability. Extensive experiments and visualizations show that MSPA significantly improves accuracy, recall, and text quality, outperforming baselines in dynamic live streaming scenarios.

cs.IR↗
arXiv2025

Timing results of 22 years for PSR J0922+0638

We conducted a timing analysis of PSR J0922+0638 (B0919+06) using data from the Nanshan 26 m radio telescope and the MeerKAT telescope, spanning from January 2001 to March 2023. During this 22-year period, we discovered a previously unreported small glitch (glitch 1) before the well-known large glitch (glitch 2), occurring at ${\rm MJD} \sim 53325(3)$, with a frequency jump amplitude of $Δν/ν\sim 0.79(6) \times 10^{-9}$. We also identified ten slow glitch events, half of which were newly detected. These slow glitches occurred quasi-periodically, with an average interval of approximately 553(21) days, fractional frequency changes ranging from $Δν/ν\sim 1.13(1) \times 10^{-9}$ to $4.08(5) \times 10^{-9}$, and a maximum fractional change in the first derivative of the frequency of $Δ\dotν/\dotν \sim -4.6 \times 10^{-3}$. Additionally, our timing noise analysis reveals a change in the spectral index for noise power before and after glitch 2, with values of $-6.0$ and $-5.3$, respectively, likely due to this large glitch. Throughout the entire observation period, the first derivative of the spin frequency ($\dotν$) showed a periodic structure. The possible modulation period was estimated to be 537(24) days before the 700-day data gap at MJD 56716 and 600(58) days afterward. We discuss the periodic oscillations in pulsar rotation as a possible manifestation of spin-down noise and quasi-periodic slow glitches.

astro-ph.HE↗
arXiv2025

BMIKE-53: Investigating Cross-Lingual Knowledge Editing with In-Context Learning

This paper introduces BMIKE-53, a comprehensive benchmark for cross-lingual in-context knowledge editing (IKE) across 53 languages, unifying three knowledge editing (KE) datasets: zsRE, CounterFact, and WikiFactDiff. Cross-lingual KE, which requires knowledge edited in one language to generalize across others while preserving unrelated knowledge, remains underexplored. To address this gap, we systematically evaluate IKE under zero-shot, one-shot, and few-shot setups, incorporating tailored metric-specific demonstrations. Our findings reveal that model scale and demonstration alignment critically govern cross-lingual IKE efficacy, with larger models and tailored demonstrations significantly improving performance. Linguistic properties, particularly script type, strongly influence performance variation across languages, with non-Latin languages underperforming due to issues like language confusion. Code and data are publicly available at: https://github.com/ercong21/MultiKnow/.

cs.CL↗
arXiv2025

Reactive Aerobatic Flight via Reinforcement Learning

Quadrotors have demonstrated remarkable versatility, yet their full aerobatic potential remains largely untapped due to inherent underactuation and the complexity of aggressive maneuvers. Traditional approaches, separating trajectory optimization and tracking control, suffer from tracking inaccuracies, computational latency, and sensitivity to initial conditions, limiting their effectiveness in dynamic, high-agility scenarios. Inspired by recent breakthroughs in data-driven methods, we propose a reinforcement learning-based framework that directly maps drone states and aerobatic intentions to control commands, eliminating modular separation to enable quadrotors to perform end-to-end policy optimization for extreme aerobatic maneuvers. To ensure efficient and stable training, we introduce an automated curriculum learning strategy that dynamically adjusts aerobatic task difficulty. Enabled by domain randomization for robust zero-shot sim-to-real transfer, our approach is validated in demanding real-world experiments, including the first demonstration of a drone autonomously performing continuous inverted flight while reactively navigating a moving gate, showcasing unprecedented agility.

cs.RO↗
arXiv2025

LangSAMP: Language-Script Aware Multilingual Pretraining

Recent multilingual pretrained language models (mPLMs) often avoid using language embeddings -- learnable vectors assigned to individual languages. However, this places a significant burden on token representations to encode all language-specific information, which may hinder language neutrality. To address this limitation, we propose Language-Script Aware Multilingual Pretraining (LangSAMP), a method that incorporates both language and script embeddings to enhance representation learning. Specifically, we integrate these embeddings into the output of the Transformer blocks before passing the final representations to the language modeling head for prediction. We apply LangSAMP to the continual pretraining of XLM-R on a highly multilingual corpus covering more than 500 languages. The resulting model consistently outperforms the baseline in zero-shot crosslingual transfer across diverse downstream tasks. Extensive analysis reveals that language and script embeddings capture language- and script-specific nuances, which benefits more language-neutral representations, proven by improved pairwise cosine similarity. In our case study, we also show that language and script embeddings can be used to select better source languages for crosslingual transfer. We make our code and models publicly available at https://github.com/cisnlp/LangSAMP.

cs.CL↗
arXiv2025

Real-Time Spatial Reasoning by Mobile Robots for Reconstruction and Navigation in Dynamic LiDAR Scenes

Our brain has an inner global positioning system which enables us to sense and navigate 3D spaces in real time. Can mobile robots replicate such a biological feat in a dynamic environment? We introduce the first spatial reasoning framework for real-time surface reconstruction and navigation that is designed for outdoor LiDAR scanning data captured by ground mobile robots and capable of handling moving objects such as pedestrians. Our reconstruction-based approach is well aligned with the critical cellular functions performed by the border vector cells (BVCs) over all layers of the medial entorhinal cortex (MEC) for surface sensing and tracking. To address the challenges arising from blurred boundaries resulting from sparse single-frame LiDAR points and outdated data due to object movements, we integrate real-time single-frame mesh reconstruction, via visibility reasoning, with robot navigation assistance through on-the-fly 3D free space determination. This enables continuous and incremental updates of the scene and free space across multiple frames. Key to our method is the utilization of line-of-sight (LoS) vectors from LiDAR, which enable real-time surface normal estimation, as well as robust and instantaneous per-voxel free space updates. We showcase two practical applications: real-time 3D scene reconstruction and autonomous outdoor robot navigation in real-world conditions. Comprehensive experiments on both synthetic and real scenes highlight our method's superiority in speed and quality over existing real-time LiDAR processing approaches.

cs.RO↗